Arav AI 🤖✨

Multi-Domain-Humanized-AI-Assistant

Arav is a highly responsive, voice-activated AI assistant with a unique Hinglish personality. Built with Python, Groq, and state-of-the-art speech technologies, Arav is designed to be a helpful, context-aware, and engaging companion.

---

🚀 Features

• Voice-Activated: Hands-free operation using wake words ("Hey Arav").
• High-Speed Responses: Powered by the Groq LPU™ Inference Engine with the Llama 3.1 model for near-instantaneous replies.
• Unique Personality: A custom "Hinglish" persona that makes interactions feel natural and fun.
• Context-Aware: Knows the current time, date, and your location for relevant answers.
• Real-time Weather: Integrated with Open-Meteo for live weather updates.
• Noise Reduction: Cleans microphone input for better recognition in noisy environments.
• Switchable Voices: Choose between male or female voice output.
• Conversation Memory: Remembers recent dialogue for follow-up questions.
• Extensible Skills:
  • Movie Recommendations
  • Finding Places
  • Emotional Support Conversations
  • Weather Forecasting
  • Fashion Tips
  • General Talk and Q&A

---

🛠️ How It Works

Arav operates on a simple yet effective loop: Listen → Process → Think → Respond

1. Wake Word Detection: Listens passively for "Hey Arav".
2. Command Recording: Starts recording after activation.
3. Audio Processing: Cleans audio via noise reduction algorithms.
4. Speech-to-Text (STT): Converts speech to text using Google STT.
5. Context Assembly: Combines the user's query, history, time, and weather.
6. LLM Inference: Sends the context to Groq for an intelligent, Hinglish-style response.
7. Text-to-Speech (TTS): Converts the response into natural-sounding audio.
8. Audio Playback: Plays the response instantly.

---

🔧 Tech Stack

• LLM Engine: Groq (Llama 3.1 8B Instant)
• STT: speech_recognition (Google STT)
• TTS: edge-tts
• Audio Processing: pyaudio, pydub, sounddevice, noisereduce
• APIs: Open-Meteo, TMDB
• Language: Python 3.9+
• Async Framework: asyncio

---

⚙️ Setup and Installation

Prerequisites

• Python 3.9+
• API Keys:
  • GroqCloud
  • TMDB (optional but recommended)
• System dependencies for PyAudio:
  • Debian/Ubuntu/Raspberry Pi OS

    sudo apt-get update && sudo apt-get install portaudio19-dev

  • macOS

    brew install portaudio

---

Installation Steps

1. Clone the repository

   git clone https://github.com/your-username/arav-ai.git
   cd arav-ai

2. Create & activate a virtual environment

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

4. Set environment variables
   Create a .env file in the root directory:

   GROQ_API_KEY="gsk_YourGroqApiKeyHere"
   TMDB_API_KEY="YourTmdbApiKeyHere"

---

📦 requirements.txt

groq
requests
speechrecognition
edge-tts
python-dotenv
pydub
numpy
sounddevice
pyaudio
scipy
noisereduce

📚 Libraries Used

Below are the libraries used throughout the project, grouped and explained for clarity.

🔧 Core Python Libraries

json

Used for saving & loading conversation history in .json format.

os

Handles environment variables, file paths, and file clean-ups.

datetime

Fetches live time info to include in the system prompt.

re

Used for regex filtering to remove emojis / markdown before TTS output.

time

Used for pauses, handling delays, and timing operations in audio processing.

🌐 Networking & API Libraries

requests

Performs external API calls including weather (Open-Meteo) and TMDB movie search.

dotenv (load_dotenv)

Loads API keys (GROQ_API_KEY, TMDB_API_KEY) from .env file securely.

🤖 AI Interaction

groq

Used to call Groq LLM (Llama 3.1) for generating conversational responses.

🎙 Audio Input / Output Libraries

speech_recognition

Handles speech-to-text using Google STT API.

pyaudio

Records microphone audio in raw PCM format.

wave

Saves raw byte-stream frames into .wav audio files.

sounddevice

Used to play final generated speech audio.

pydub

Converts MP3 → WAV, handles buffering, helps playback preparation.

edge_tts

Generates Text-to-Speech using Microsoft Edge Neural TTS voices.

io

Handles in-memory audio buffers for TTS playback.

🔊 Audio Processing & Noise Reduction

numpy

Performs array transformations for audio analysis & TTS playback.

scipy.io.wavfile

Used to read/write numeric WAV data for denoising.

noisereduce

Applies deep learning–based noise reduction on recorded voice commands.

🏠 Hardware Interaction

RPi.GPIO

Controls Raspberry Pi GPIO pins (optional future extension: LED, button, sensor triggers).

📦 Project Overview

This project is a fully offline + online hybrid voice assistant, named Arav, powered by:

Groq Llama 3.1 model

Google Speech Recognition

Microsoft Edge Neural TTS

Noise-reduced microphone input

Wake-word detection

Location-aware weather

Conversational memory

⚙️ Core Architecture Summary

1. Wake-Word Listener

Continuously records 3-second audio chunks and checks if the user said:

["arav", "ara", "aarav", "naru", "hey naru"]

2. Voice Command Recording

Once awakened, it records a 7-second command:

Applies noise reduction

Converts to WAV

Sends to Google STT

3. Command Understanding

Processes:

Voice switching

Location change

Sleep mode

Movie queries

Weather questions

4. LLM Interaction

Forms a system prompt including:

Time

Weather

Location

Personality profile (Hinglish, witty, meme-ish)

Movie recommendation

Then calls Groq API for reply.

5. Speech Output

Cleans text of emojis & markdown → sends to Edge TTS → plays via sounddevice.

📁 Important Configurations
Audio Settings:
MIC_SAMPLE_RATE = 16000
MIC_CHANNELS = 1
MIC_FORMAT = pyaudio.paInt16
MIC_CHUNK_SIZE = 1024

Filenames:

temp_raw_audio.wav  
temp_clean_audio.wav  
conversation_history.json

🔥 Why this Architecture Works Great

✔ Lightweight
✔ Super fast (Groq inference)
✔ Good noise reduction
✔ Real conversational memory
✔ Wake-word based hands-free experience
✔ Human-like Hinglish personality

🚀 Running the App

1. Install Dependencies

pip install -r requirements.txt

2. Add your .env file:

GROQ_API_KEY=your_key
TMDB_API_KEY=your_key

3. Run:

python main.py

Arav waits for your voice: “hey Arav”

→ processes your command → responds naturally.